Innovation Cognitive Science

Why New Ideas Are Built From Known Concepts

Existing knowledge shapes imagination, scientific work, and new learning. The evidence supports building on what we know, while leaving room to question the frameworks we inherit.

By

Part 2 of 4 in the series “The Background Work of Innovation.”

Familiar clay shapes combined into a new sculptural construction

We often talk about innovation as though it means escaping what we already know. Think outside the box. Start with a blank page. Forget how it has always been done.

In the first article in this series, I examined whether breaks, rest, and sleep can improve later problem solving. Sometimes they do, although the mechanism and the necessary conditions remain unsettled. That raises the question this article addresses: what does the brain build new ideas from?

My hypothesis is that it starts from known concepts, which expand into something new. The research supports that part of the hypothesis more strongly than the popular picture of innovation would suggest. Studies of imagination show that invented ideas closely follow the structure of familiar categories.[1] An analysis of 17.9 million scientific papers found that the highest-impact work was grounded mostly in conventional combinations of prior knowledge.[2] And animal research shows that an existing framework of knowledge can make new information much faster to absorb.[3]

What the evidence does reveal is a useful correction to the blank-page picture: new ideas are rarely built from nothing. They are built from what the mind already has, and what the mind has shapes what it can build.

Imagination follows the shape of what we know #

In a 1994 paper in Cognitive Psychology, Thomas Ward asked college students to imagine animals that might live on another planet. The task invited novelty, but its starting category, animals, was already familiar.[1]

Most of the imagined creatures had the properties of typical animals on Earth: bilateral symmetry, sensory organs, and appendages.[1] Ward called this structured imagination. When people create a new member of a familiar category, the properties of that category shape what they create.

Ward also found that instructions and task constraints could move participants from particular Earth-animal examples toward broader knowledge frameworks.[1] That matters because the influence of prior knowledge was flexible, not an inescapable copying rule.

One way to describe the possibility is conceptual expansion: extending a familiar concept beyond its usual instances. In this series, that is a candidate process, not a claim that every new idea must emerge through the same sequence.

These tasks show that familiar categories shape some new creations. They cannot prove that every act of imagination starts with a fully formed concept, or that all innovation is recombination. Prior knowledge influences invention without determining every outcome.

The highest-impact work is mostly conventional #

Laboratory tasks raise an obvious question. Does the same pattern appear in real innovation?

In a 2013 study in Science, Brian Uzzi and colleagues analyzed 17.9 million research papers indexed in the Web of Science. They measured how often journals appeared together in reference lists relative to a randomized baseline. Those journal pairings were a proxy for combinations of knowledge, not a direct reading of the concepts in each paper.[2] Their abstract opens with the observation that:

“the building blocks of new ideas are often embodied in existing knowledge.” [2]

Papers combining a highly conventional reference pattern with some unusual pairings were more than twice as likely to become highly cited as papers lacking that combination. In this dataset, teams were 37.7% more likely than solo authors to introduce novel combinations into conventional backgrounds.[2]

This was an observational association. Citation impact is not the same as originality, correctness, or usefulness, and the study did not manipulate how scientists thought. It supports a relationship between citation patterns and later impact in this corpus. It does not establish a causal recipe for an innovative mind.

Known structure makes new learning faster #

A related question is whether organized prior knowledge changes how quickly new information can be learned. A 2007 study in Science by Dorothy Tse, Richard Morris, and colleagues tested this in rats.[3]

The rats were first trained over several weeks to learn which flavors of food could be found at which locations in an arena. As training progressed, they appeared to develop what the researchers called a schema: an organized framework of associations. The researchers then introduced new flavor-location pairs, each trained only once.[3]

With the schema in place, new associations could be learned in a single trial. Neurotoxic lesions of the hippocampus made 48 hours after learning spared later recall, whereas lesions made after three hours disrupted recall of the new pairs. The researchers inferred unusually rapid consolidation when new material fitted an established schema. The experiment measured later memory performance, not whether a memory had literally been erased.[3]

This is animal research on flavor-location memory. A learned spatial schema is not identical to a human concept or a hierarchy of expert principles. The result makes a narrower point with biological evidence: an established framework can change how new, compatible information is learned and retained. Extending that finding to human innovation requires another step.

Offline processing needs raw material #

This connects back to the first article in this series. In the 2004 sleep study by Ullrich Wagner and colleagues, sleep more than doubled the number of participants who discovered a hidden rule in a number task. But sleep did not help participants who had not first trained on the task.[4]

That detail fits my hypothesis: prior learning mattered in this experiment. But recently practiced task representations are not necessarily the same thing as general concepts. The connection is plausible, not a demonstrated identity between memory consolidation and conceptual innovation.

For practical work, that makes the quality of preparation worth examining alongside the break itself. Knowing more, organizing what is known, and selecting the relevant knowledge are distinct possibilities to test.

Known concepts are a starting point, not a limit #

None of this means innovation is simply recombination. The same research that shows the power of existing knowledge also shows its pull. Ward's participants tended to reproduce the features of Earth animals, and novel combinations were rare in the scientific literature Uzzi's team analyzed.[1][2]

A hypothetical example makes the distinction concrete. Consider an operations engineer asked to design a new way to schedule batch jobs. She has never built a scheduler. But she has handled dozens of incidents in which retries piled onto an overloaded service, queues backed up, and timeouts triggered still more retries.

She might draw on those incidents when designing the scheduler. The example illustrates how prior experience could contribute. It is not evidence about what every engineer must think first.

The question is whether she stays with the incidents themselves or builds something above them: a pattern she can apply to a system she has never seen. That is the subject of the next article.

The real test is what the concepts can support #

This part of the hypothesis makes a testable prediction. If new ideas are built from known concepts, then what people know, and how that knowledge is organized, should shape the ideas they generate.

A stronger test would randomly teach different groups the same facts through different relationships or examples, then give them the same new problem. Prior expertise and learning time would need to be measured or matched. Comparing existing experts with novices alone would leave experience, selection, and many other differences entangled.

Independent judges, unaware of the training condition, could rate novelty and usefulness separately. Instead of calling ideas “entirely new,” the study would define which relationships were absent from training and test whether people used those relationships successfully.

The scaffolding claim needs a separate comparison: teach an organized framework or equally detailed facts, then introduce compatible and conflicting information. A framework that helps only with compatible information would support a conditional learning benefit. Failure on conflicting material could reveal the cost of the same organization.

Innovation extends what we know #

The research challenges a simple blank-page picture. Familiar categories shape invented creatures, citation patterns relate to scientific impact, and an established schema can accelerate new memory formation. These are three different findings that make prior knowledge central to a useful account of innovation.

For my hypothesis, they provide converging support for taking known concepts seriously. They do not yet show that category use, citation recombination, and memory consolidation are one process, or that they account for every source of novelty.

The most important question is not “How do we escape what we know?” It is “What do we know well enough to build on?”

Previous in this series: Part 1: What the Brain Does After We Stop Working on a Problem

Next in this series: Part 3: How Known Concepts Become Layers of Structure

References #

  1. Ward, T. B. (1994). Structured imagination: The role of category structure in exemplar generation. Cognitive Psychology, 27(1), 1–40.

  2. Uzzi, B., Mukherjee, S., Stringer, M., & Jones, B. (2013). Atypical combinations and scientific impact. Science, 342(6157), 468–472.

  3. Tse, D., Langston, R. F., Kakeyama, M., Bethus, I., Spooner, P. A., Wood, E. R., Witter, M. P., & Morris, R. G. M. (2007). Schemas and memory consolidation. Science, 316(5821), 76–82.

  4. Wagner, U., Gais, S., Haider, H., Verleger, R., & Born, J. (2004). Sleep inspires insight. Nature, 427(6972), 352–355.

Cite this article #

When quoting or referencing this article, credit Paul Greenberg and RapidKT and link to the article. Include the section title when directing readers to a particular passage.

Suggested citation

Greenberg, P. (2026, September 30). Why New Ideas Are Built From Known Concepts. RapidKT. https://rapidkt.com/pages/blog/why_new_ideas_are_built_from_known_concepts

In-text reference: (Greenberg, 2026).

RapidKT Plain Text Website